ECCV paper: one scalar per patch from pre-trained ViTs enables fast real-world robot navigation
chriswolfvision · x · 2026-09-12
Christian Wolf's team presents an ECCV study on real-world robot navigation: visual encoders distilled from heterogeneous teachers can be bottlenecked to just one scalar per patch via attention projection, yet still support fast moving navigation in a real building. Policies are pre-trained with privileged Lidar input and then fine-tuned to RGB-only. An interpretable affordance-linked structure emerges. A large-scale 966-episode / 24km real-robot evaluation is covered in a companion post.
More from Embodied
- Dev builds real 'sidequest' device that dances only to music it likes — wavefnx · 2026-10-31
- Indie dev prototypes custom secure hardware in a day with Rust firmware and root-key isolation — wavefnx · 2026-10-31
- UCLA's ChairNav teaches autonomous wheelchairs to learn each rider's preferences — zhoubolei · 2026-10-31
- Amazon Echo kills local voice storage: everything goes to the cloud from March 28 — alex_verem · 2026-10-31
- HP paid $116M for Humane, then bricked every $699 AI Pin 10 days later — alex_verem · 2026-10-31
- Moxie, the $800 AI robot friend for kids, went silent overnight after shutdown — alex_verem · 2026-10-31